Leaked Google Doc
Few people (except everyone into IM) realise that you can read the Google search patent publicly. Yes, Google it! Of course the patent (and there is more than one BTW) merely describes the mechanism, NOT how the elements of the algorithm are subtly changed multiple times during an unknown period.
So, apart from that being pretty interesting to those that have not heard of it before, why should it be interesting to you?
Well, most people understand that Google took over Yahoo’s dominance for one reason. Automation. Yahoo = people power. Google = server power. But surprise surprise, the only way to check the ‘algorithm’ is to use people to see if the search results make sense.
So, these people need some training then, right? Oh yes. And every now and again someone leaks the training manual, which has just happened again in October [2011 in case you are reading this from an Apple TimePod tm].
So how does it work. Simple. Human readers get fed a bunch of sites to see if they match up to the search criteria that led to them being served up to the user. The BIG problem of course is one of subjectivity (how can a mere mortal ever replace a computer). Porn is the easy one (for fetishes of this decade at least), but what about relevancy?
Let’s take a hypothetical case of a relevant site over another equally relevant site. And let’s look at the search term ‘cloud’. The guidelines start with the term ‘Vital’ for grading results. So what is vital about ‘cloud’. Well of course it is the cloud of the internet (what did you expect!) in first place. Followed by a real cloud. Then a couple more virtual clouds, then some cumulo-nimbus, etc.
This is great and it shows Google as its best. But if you dig deeper you will find some well known search terms are a little cloudy. It is of course all about relevance and I an happy with that, but the point here is two fold.
If you add a human into the algorithm, you bring a bunch of trouble. If you then try and cure the trouble with a set of rules, you will only make the trouble worse. Humans are not machines QED.
So, what’s the point of this? If you want to know how Google *might* review your site, you need to know the criteria of the human aspect. So take a look at this excellent article from SearchEngineLand. It is worth your time. This is not so much a leaked Google doc as a completely public one!